The present disclosure is directed to techniques for signal modulation, more particularly, modifying modulation signals to avoid feature extraction from digital speech data.